Solid tumors require new blood vessels to grow beyond a few cubic millimeters. Tumor blood vessels are complex and dysfunctional and multiple cell types may coalesce to form the tumor vasculature. In a striking example, some tumor cells may directly integrate within vascular structures or form tumor cell-lined conduits that carry blood and fluid (termed vascular mimicry, VM). We have recently isolated and characterized a novel subpopulation of vascular-like tumor cells identified by expression of the vascular cell adhesion molecule, CD31. These CD31+ tumor cells do not express VEGF receptors, they down-regulate the neural crest transcription factor/CD31 repressor AP-2alpha, and they form tumor blood vessels when engrafted in mice. Furthermore, CD31+ tumor cells do not respond to VEGF inhibition and are enriched in tumors challenged with VEGF neutralizing antibodies. In aim 1 we will use clonal populations of CD31- and CD31+ tumor cells we have derived from human and mouse cell lines and a spontaneous mouse melanoma model to determine the functional role of CD31 in forming perfused vascular structures in tumors. Aim 2 is to use cell ablation strategies and cutting-edge tumor perfusion studies to determine how CD31+ tumor cells mediate escape from anti-angiogenic therapy. In aim 3 we will define how loss of the neural crest specifer AP-2alpha controls CD31 expression and generates VM-competent tumor cells. For this aim, we will use genetic deletion and over- expression studies and intravital microscopy to visualize VM-competent tumor cells implanted in the mouse ear. Upon completion of our study aims, we will clarify how CD31+ tumor cells form functional connections with the host vasculature, the molecular mechanisms that generate and maintain this unique subpopulation, and how these vascular like tumor cells mediate escape from anti-angiogenic therapy.